An element using a repeating microscopic structure of p-type and n-type layers wherein an electric field relaxation phenomenon called the RESURF (REduced SURface Field) effect is applied in place of the uniform n-type drift layer of a conventional MOS-FET (Metal Oxide Semiconductor-Field Effect Transistor) has been proposed in, for example, U.S. Pat. No. 6,040,600. In this element a low ON resistance is obtained in the ON condition due to the n-type drift layer of which the impurity concentration is higher than the concentration of the uniform n drift layer in the conventional structure by approximately one order while in the OFF condition the entire electric field is relaxed due to a three-dimensional multiple RESURF effect of nip layers. Thereby, a withstand voltage several times as large as the main withstand voltage conventionally obtained by a high concentration single n-type drift layer alone can be implemented and, in principle, an STM (Super Trench power MOS-FET) structure that can obtain a value lower than the Si limitation (Ron, sp=5.93×10−9 BV2.5, wherein specific resistance is proportional to the main withstand voltage to the power of 2.5) wherein the relationship between the main withstand voltage and the specific ON resistance is limited can be obtained.
In an actual element, however, this repeating microscopic structure of p-type and n-type layers cannot be repeated infinitely in an edge portion of the chip and, therefore, there is a problem wherein a drop in the main withstand voltage is great in a “termination” portion of a termination structure where the repetition ends. In the following, a prior art and problem thereof are described from such a point of view.
FIG. 148 is a cross sectional view schematically showing the first configuration of a semiconductor device according to a prior art and shows a configuration that corresponds to a case where a MOS-FET is posited as a concrete active element structure. In reference to FIG. 148, an n− epitaxial layer 102 is formed on the first main surface side of an n+ drain region 101 of the MOS-FET. A pn-repeating structure wherein n-type drift regions  103 and p-type impurity regions 104 are repeated in alternation is formed within this n− epitaxial layer 102.
Here, though the vicinity of the center of this element having the pn-repeating structure is omitted for the purpose of simplification of the description, conventionally a combination of several hundreds to several tens of thousands of repeated pairs of n-type drift regions 103 and p-type impurity regions 104 exists in this portion. The n-type impurity concentration of n-type drift region 103 and the p-type impurity concentration of p-type impurity region 104 in each pair are set at substantially the same level.
A p-type body region 105 is formed on the first main surface side of p-type impurity region 104. This p-type body region 105 is also located on, at least, a portion of n-type drift region 103 on the first main surface side so as to form a main pn junction with n-type drift region 103. An n+ source region 106 of a MOS-FET and a p+ contact region 107 for making a low resistance contact with p-type body region 105 are formed side by side in the first main surface within this p-type body region 105.
A gate electrode 109 is formed above the first main surface so as to face p-type body region 105 located between n-type drift region 103 and n+ source region 106 via a gate insulating film 108. When a positive voltage is applied to this gate electrode 109, p-type body region 105, which faces gate electrode 109, is inverted to an n-type so that a channel region is formed.
A source electrode 110 made of a material including aluminum (Al), for example, is formed on the first main surface so as to be electrically connected to n+ source region 106 and p+ contact region 107.
A drain metal wire 111 is formed on the second main surface so as to contact n+ drain region 101.
Here, in the actual element, the source electrode part is electrically connected to n+ source region 106 and p+ contact region 107 through a contact hole provided in an interlayer insulating film on the first main surface and via a barrier metal. In the present application, however, this portion is not important and, therefore, the source electrode part is simplified and expressed using solid lines throughout all of the drawings.
In addition, though n+ drain region 101 is several times to several tens of times thicker than the effective element portion in the actual element, n+ drain region 101 is expressed as thinner than the effective element portion in the drawings for the purpose of simplification. In addition to the above, scales, ratios of dimensions, and the like, are deformed in order to simplify the expression and, therefore, the respective dimensions in the drawings are not necessarily precise.
A multiple guard ring structure made of p-type impurity regions 115, for example, is provided as a termination structure of the pn-repeating structure.
In this configuration, n-type drift regions 103 and p-type impurity regions 104, respectively, have substantially the same impurity concentration in the center portion and edge portions of the pn-repeating structure.
FIG. 149 is a cross sectional view schematically showing the second configuration of the semiconductor device according to the prior art. In reference to FIG. 149, an n− epitaxial layer 102 has a buried multi-layer epitaxial structure and a p-type impurity region 104 is formed of a plurality of p-type regions 104a that are integrated in the depth direction of the semiconductor substrate in this configuration. In this configuration, p-type impurity regions 104, respectively, have the same impurity concentration in the center portion and edge portions of the pn-repeating structure.
Here, the concentration distribution in the upward and downward directions of each p-type impurity region 104 is an intrinsic structure and this is a concentration distribution due to the manufacturing method, which has no bearing on the concentration gradient in the part in the lateral direction discussed in the present invention. In addition, though in the drawing the concentration gradient in the upward and downward directions is depicted in only two stages for the purpose of simplification, in practice this concentration sequentially changes.
A manufacturing method according to this prior art is characterized in that n− epitaxial layer 102, having a comparatively high concentration to the extent that the concentration thereof is balanced with that of the p-type layers, is used for the purpose of simplifying the process of formation of the buried layers. A heat treatment is carried out after forming p-type buried diffusion layers 104 a within n− epitaxial layer 102 in such a manner and, therefore, p-type impurity region 104 becomes of a form well-known in Japan as “round sweet balls of confectionary on a skewer.”
FIG. 150 is a cross sectional view schematically showing the third configuration of the semiconductor device according to the prior art. In reference to FIG. 150, n-type drift regions 103 and p-type impurity regions  104 form pairs and a trench 123 filled in with a filling 124 is arranged between the members of each combined pn pair in this configuration.
FIG. 151 shows the appearance of electrical field concentration in the structure corresponding to this FIG. 150. The dark portion in this figure indicates a portion of high electrical field concentration and it is seen that an electrical field concentrates on portions (regions shown by arrows) wherein the pn-repeating structure ends.
Here, in this FIG. 151, an FP (Field Plate) structure is adopted for the termination structure portions instead of the multiple guard ring called an FLR (Field Limiting Ring) or an FFR (Floating Field Ring).
Here, the other parts of the above described configurations shown in FIGS. 149 and 150 are approximately the same as in the configuration shown in FIG. 148 and, therefore, the same symbols are attached to the same members, of which the descriptions are omitted.
As described above, according to the first to third prior arts there are structures wherein conventional termination structures such as a guard ring, an LFR, a JTE (Junction Termination Extension) and an FP are combined in the portions wherein pn-repeating structures end. By combining such termination structures, however, only a withstand voltage far lower than the high withstand voltage obtained within the cell in the center portion of the pn-repeating structures can be obtained in portions wherein the pn-repeating structure ends. Therefore, though the element operates, there is a problem wherein the trade-off relationship between the main withstand voltage and the ON resistance does not improve.
In addition, the content of the following Prior Art 1 has been announced as a method for preventing the loss of the high withstand voltage of the main cell portion by setting a specific concentration of the p-type layers and of the n-type layers outside of the portions wherein the pn-repeating structures ends. According to this technique, however, there is a problem wherein implementation is difficult due to the reasons described below.
The above described Prior Art 1 is described in “Junction Termination Technique for Super Junction Devices” that was announced in, for example, ISPSD 2000 (International Symposium on Power Semiconductor Devices & ICs) of CPES (Center for Power Electronics Systems), Virginia Polytechnic Institute and State University.
This Prior Art 1 shows improvement of the termination structure itself in the pn-repeating structure.
In addition, the structure shown in FIG. 152 is shown in the above described Prior Art 1. In reference to FIG. 152, a region of which the effective conductive type and concentration can be regarded as those of a low concentration p− region in a fan form of a quarter of a circle having a radius of R of the thickness (depth) of an n layer is formed from a portion wherein the repetition of p layers 204 and n layers 203 ends. However, a p− region cannot actually be formed to have such a concentration distribution. Therefore, it is necessary for the concentration distribution of the effective p− region to have an attenuation curve as shown in FIG. 153.
In order to implement this, a configuration is used wherein the concentration and the width of n-type regions 203 are constant while the concentration of p-type regions 204 is constant and the widths thereof are changed such as in the SJT (Super Junction Termination) structure shown in FIG. 154. Thereby, the same effects as of the changing of the effective concentration can be obtained according to the description of Prior Art 1.
In addition, the only requirements at this time are a form wherein the equipotential surfaces are aligned in fans at equal intervals as shown in FIG. 155 and a zigzag electrical field intensity distribution that is exposed to the surface wherein the peaks and the troughs have the same height and depth, respectively.
In addition, in this Prior Art 1 each of the concentrations of pi regions 204 and ni regions 203 are posited as being uniform within the single diffusion layer in the upward, downward, leftward and rightward directions. There is a problem, however, wherein the original effects of Prior Art 1 cannot easily be exercised when the formula for the relationship of the pn concentration ratio is not fulfilled in the case that the absolute values of the concentration greatly change or when the description of such a relationship becomes extremely complex so that the precision of proximity is reduced.
Concretely, there is a description that “along the SJT surface, . . . in the following calculation.” in right column of page 2 to the left column of page 3 in the main body of Prior Art 1. In this description the volume represented by the concentration and the width of each portion may be set so as to satisfy equation (5) in Prior Art 1 so that the electrical field distribution closest to the surface does not reach to the critical breakdown electrical field.
In other words, this Prior Art 1 discloses the design of the entirety of the element in a form that includes the termination structure by literally extending the super junction structure of the repeating cell portions to the termination structure portions in some manner according to SJT, that is to say, “Super Junction Termination structure,” wherein a repeating cell portion in the center and a termination structure have a one-to-one correspondence so as to be indivisible having a very limitative structure while “manner of connection” of a repeating cell portion to a general termination structure portion is described in the present invention, which is essentially different from the above.
In the case that the distribution required for the p-type acceptor concentration distribution in the moving radius direction in FIG. 153 is formed according to the repetition of pn layers, the electrical field distribution closest to the surface becomes of a zigzag form and, in the case that the peaks and troughs all have the same value, the maximum withstand voltage can be obtained. Therefore, in the case that the all of the concentrations of n and p regions 203 and 204 are made uniform so that the equipotential surfaces (lines) distribute in fan forms at equal intervals, as shown in FIG. 155, it is necessary to carry out an adjustment of the width of each of the regions 203 and 204.
In addition, SJT is considered to be impractical because it has the following two problems.
First, concentration regulation for forming an SJT structure is too complicated and it is necessary to apply an interval design that agrees with the concentration arrangement of the repeating cell portions that are different from the termination structure portions to the SJT part after examining the arrangement in detail before carrying out the actual design and, in addition, it is physically and mechanically difficult to fabricate a semiconductor chip structure to include terminal edges. On the other hand, the present invention has the advantage that both design and manufacturing method are simple because the relative concentrations in the vicinity of the terminal edges of the repeating cell portions may be adjusted using comparatively simple arithmetic.
Secondly, an SJT structure can only be implemented in the case of manufacture by means of a buried multi-layer epitaxial growth method and lacks versatility in that it cannot be actually manufactured in the case wherein a trench sidewall diffusion is used.
Furthermore, as described in the main body of Prior Art 1, there is a problem wherein this technique lacks versatility in that it is impossible to apply this technique in an element structure wherein a trench system is applied due to restrictions of the manufacturing technology even though the application to a multi-layer epitaxial structure is, in principle, possible.
Next, the technology disclosed in U.S. Pat. No. 5,438,215 is described as prior art 2 in FIG. 156.
In reference to FIG. 156, a vertical-type MOS-FET has inside region  301 that is doped so as to be a low level n-type. A base region 303 of the opposite conductive type (p) is provided in the upper side surface 302 of the semiconductor substrate. A source region 304 of the first conductive type (n) is buried within base region 303. A gate electrode 308 is arranged above surface 302 so as to be insulated from the surface. A drain region  307 that is highly doped so as to be of the same conductive type as inside region 301 is provided in the surface 306 on the opposite side.
Auxiliary semiconductor regions 311 and 312 are arranged in a range of the space-charge region that spreads at the time of reverse voltage application within the inside region 301. At least two regions 211 of a conductive type opposite to that of the inside region are provided. Auxiliary regions 312 having the same the same conductive type (n) as inside region 301 and being more highly doped than the inside region arranged between regions 311. The auxiliary regions are surrounded from all directions by a single region. This single region is of the same conductive type as the inside region, as well as regions 312, and is more highly doped than the inside region.
Though in this configuration a portion, wherein an active cell is formed, is buried in n− region 301, which has a low concentration, the impurity concentration of this outer peripheral portion is not specifically described and only the method of formation of a cell portion is discussed.
In addition, in general the impurity concentration of a portion wherein a pn-repeating structure is not formed in this Prior Art 2 is presumed to be set at the impurity concentration that is reverse calculated from a value obtained by adding a manufacturing margin to the element withstand voltage set for the power MOS-FET of a conventional structure (structure that does not have pn repetition). However, that leads the electrical field distribution in the termination structure portions in the pn-repeating structure to become triangular so as to differ from an electrical field distribution in a trapezoidal form that is implemented in the cell portion. Therefore, in the same manner as in the above described Prior Art 1, the difference in the electrical field distribution between the inside of repeating cells and the termination structure portions becomes greater so that there is a problem wherein a high withstand voltage, which is essentially obtained in a cell portion, cannot be implemented although the relationship between the main withstanding voltage and the ON resistance is improved in comparison with the conventional MOS-FET structure.